Lisbon, Portugal
Development Services
Contact us
ENG
Follow us
How we built a field sensor network, edge processing layer, and agronomist dashboard that replaced blanket fertilizer application with real-time, field-specific nutrient recommendations — reducing input costs while improving yield consistency across a multi-thousand hectare operation.
Multi-parameter LoRaWAN nodes — NPK, moisture, temperature, EC, pH — across hundreds of fields.
Crop type, growth stage, historical yield, and satellite-derived NDVI imagery for field-level calibration.
Real-time weather forecast and NVZ nitrogen limits for the field's regulatory zone.
Integrates sensors, agronomic context, weather, and regulatory limits to produce variable-rate NPK prescriptions per management zone — with confidence intervals tied to sensor coverage.
The client operates a large-scale arable farm growing cereals and oilseeds across varied terrain and soil types, with an annual fertilizer budget running into seven figures. Fertilizer was applied on a schedule driven by crop growth stage and historical averages — the same rate to every field, on the same date, regardless of what was actually in the soil.
Fields with adequate residual nitrogen were receiving the same application as depleted fields. Leaching losses from over-application were a regulatory and environmental liability as nitrate buffer zone (NVZ) compliance requirements tightened. They needed a platform that could tell them what was in each field, and what to put where.
The core challenge was not just data collection — it was building a system that could translate raw sensor readings into agronomically valid, field-level fertilizer recommendations that a farm manager could act on confidently. The platform needed to be both the measurement layer and the decision-support layer simultaneously, starting from zero with no prototype to extend.
Deploying a dense sensor network across hundreds of fields, in variable weather and minimal connectivity, required hardware and edge processing decisions that standard IoT deployments don't face.
Soil NPK readings alone don't produce a fertilizer recommendation. The AI dosage engine needed to integrate sensor data with crop type, growth stage, weather forecast, and regulatory limits.
A recommendation in a dashboard is only half the solution. The prescription map needed to reach the spreader in a format the client's equipment could execute autonomously across field zones.
Many fields had no mobile signal or WiFi. The platform needed an edge architecture that could buffer data locally, process recommendations offline, and sync when connectivity was available.
We deployed a multi-parameter soil sensor network across the operation, measuring soil moisture, temperature, electrical conductivity, pH, and ambient conditions. Sensors communicate via LoRaWAN to field-edge gateways that aggregate readings and buffer data during connectivity gaps.
Sensor data is combined with historical soil survey data, satellite-derived NDVI imagery, and elevation models to build a continuous soil variability map for each field. Zone boundaries are updated dynamically as new sensor data arrives.
The AI recommendation engine integrates soil sensor readings with crop type, growth stage, historical yield data, weather forecast, soil texture, and NVZ regulatory limits to generate a variable-rate fertilizer prescription for each management zone.
The dashboard presents all active fields in an interactive map view, colour-coded by current nutrient status. Agronomists can review, adjust, and approve prescriptions before release, with economic and environmental impact shown for each decision.
Approved prescriptions are exported as variable-rate application files in ISOXML format. The spreader reads the prescription map and adjusts application rate automatically as it moves between management zones. GPS-tracked application data is imported back into the platform after each run.
Every fertilizer application is automatically logged — field, date, product, rate, GPS coverage map — building the documented Nutrient Management Plan required under NVZ regulations. The platform stores 5 years of application history for farm assurance audits and lender sustainability reporting.
Multi-parameter LoRaWAN nodes with edge gateways and offline buffering for low-connectivity field environments.
Dynamic management zones combining sensor data, NDVI imagery, soil surveys, and elevation modelling.
Variable-rate NPK recommendations integrating soil sensors, crop stage, weather forecast, and NVZ regulatory limits.
Interactive field map with real-time nutrient status, prescription review, economic impact, and approval workflow.
VRA prescription export in ISOXML for autonomous variable-rate application with GPS-tracked import after each run.
Flutter app for agronomists with sensor status, field alerts, prescription approval, and spreading-job management in the field.
Satellite imagery ingestion for crop health overlay on field maps, calibrating sensor readings against aerial crop status.
Automated NMP documentation in regulatory format, field nitrogen balance, application logs, and 5-year audit trail.
Real-time flagging of prescriptions approaching or exceeding NVZ nitrogen limits, with compliance margin calculations.
Eliminating over-application in well-stocked zones produced a 23% reduction in total fertilizer spend in the first full season — the single largest input saving in the operation's history.
Targeted application in previously under-nourished zones delivered a measurable yield uplift, particularly in variable-texture fields where rate variation was most significant.
The network covers the full operation with sufficient density to capture within-field variation meaningful enough to drive zone-level prescription differences.
Every fertilizer decision is automatically logged. The Nutrient Management Plan generates itself from application records, removing weeks of manual administration each season.
Built for sensor scale, agronomic accuracy, and offline-first operation. Every layer of the stack was chosen for its track record in field deployments where connectivity isn't guaranteed and uptime matters.
If you're building a precision agriculture platform, IoT sensor network, or AI decision engine for agricultural operations, we've delivered sensor networks, AI recommendation engines, and precision application integrations for large-scale operations. It's worth a chat.